Improve Space descriptions

README.md

@@ -23,3 +23,9 @@ One-stop lab for exploring spectrograms, LWM embeddings, and lightweight evaluat
 1. Select the **Spectrograms** and **t-SNE Analysis** tabs to explore embeddings.
 2. Switch to **Modulation Classification** or **Joint SNR/Doppler Evaluation** to run the k-NN prototype with adjustable train/test splits.
 3. Provide custom data (optional) to benchmark against bundled samples.
+
+## Tab Cheat Sheet
+- **Spectrograms:** Inspect raw Sub-6 GHz spectrograms per technology/SNR/modulation/mobility before feature extraction.
+- **t-SNE Analysis:** Recreate the "colored by SNR vs. modulation" comparisons from `plot/plot_tsne.py` with balanced sampling.
+- **Modulation Classification:** Benchmark a lightweight k-NN head (standing in for the residual 1D-CNN) on LWM embeddings vs. raw inputs.
+- **Joint SNR/Doppler Evaluation:** Compare MoE embeddings and raw spectrograms on the 14-way SNR/mobility task that feeds the beam selector.

app.py

@@ -749,10 +749,26 @@ with gr.Blocks(title="LWM-Spectro Lab") as demo:
     with a lightweight k-NN prototype classifier for joint SNR/Doppler recognition.
     """
     )
+    gr.Markdown(
+        """
+    📡 **Sub-6 GHz → mmWave Beam Prediction Task**
+    - LWM extracts signal-centric features from Sub-6 GHz spectrograms, mirroring the first stage of the beam prediction pipeline.
+    - A lightweight downstream model (residual 1D-CNN, ≈500k params) consumes those embeddings to rank mmWave beams from the codebook.
+    - The bundled dataset mixes six DeepMIMO scenarios that were held out from LWM pre-training, highlighting cross-scenario generalization.
+    - Use the tabs below to inspect raw inputs, visualize the embedding space, and benchmark simple classifiers before deploying larger heads.
+    """
+    )
 
     with gr.Tabs():
         with gr.Tab("Spectrograms"):
-            gr.Markdown("Visualize raw 128×128 spectrograms with optional filters.")
+            gr.Markdown(
+                """
+            ### 🔎 Spectrogram Studio
+            - Peek at the raw 128×128 Sub-6 GHz spectrograms that drive the beam prediction task.
+            - Filter by technology/SNR/modulation/mobility to understand how diverse the training pool is across scenarios.
+            - Use the gallery to sanity-check preprocessing before sending samples through LWM or downstream models.
+            """
+            )
             with gr.Row():
                 with gr.Column(scale=1, min_width=320):
                     spec_tech = gr.Dropdown(
@@ -781,6 +797,14 @@ with gr.Blocks(title="LWM-Spectro Lab") as demo:
             spec_tech.change(update_modulation_choices, inputs=spec_tech, outputs=spec_mod)
 
         with gr.Tab("t-SNE Analysis"):
+            gr.Markdown(
+                """
+            ### 🌀 Embedding vs. Raw Space
+            - Run quick t-SNE sweeps on either LWM embeddings or raw spectrogram vectors.
+            - Toggle **Color By** to mirror the "colored by modulation vs. SNR" comparisons from the CLI examples.
+            - Balanced per-SNR sampling plus configurable perplexity help match the figures you generate locally with `plot/plot_tsne.py`.
+            """
+            )
             with gr.Row():
                 with gr.Column(scale=1, min_width=300):
                     gr.Markdown("### Filters")
@@ -865,7 +889,14 @@ with gr.Blocks(title="LWM-Spectro Lab") as demo:
             )
 
         with gr.Tab("Modulation Classification"):
-            gr.Markdown("Compare LWM embeddings vs raw spectrograms for per-technology modulation classification.")
+            gr.Markdown(
+                """
+            ### 🎯 Lightweight Modulation Head
+            - Prototype how well the frozen LWM backbone separates modulation formats for each technology.
+            - The adaptive k-NN classifier approximates the behavior of the downstream residual 1D-CNN before heavy training.
+            - Sweep train/test splits and seeds to gauge robustness when only a portion of DeepMIMO-style data is labeled.
+            """
+            )
             with gr.Row():
                 with gr.Column(scale=1, min_width=320):
                     mod_tech = gr.Dropdown(
@@ -889,6 +920,14 @@ with gr.Blocks(title="LWM-Spectro Lab") as demo:
             )
 
         with gr.Tab("Joint SNR/Doppler Evaluation"):
+            gr.Markdown(
+                """
+            ### 🌪️ Joint Channel Dynamics Benchmark
+            - Evaluate the precomputed MoE embeddings on the 14-class joint SNR/Doppler recognition task.
+            - Mirrors the second stage of the beam prediction pipeline where mobility-aware cues guide codebook pruning.
+            - Upload or reference Hub-hosted tensors to compare MoE vs. raw spectrogram baselines before fine-tuning heavier heads.
+            """
+            )
             if evaluation_disabled:
                 gr.Markdown(
                     "⚠️ Precomputed MoE embeddings are not bundled in this Space build. Upload a dataset locally to run evaluations."